Somatic symptoms in adolescence as a predictor of severe mental illness in adulthood: a long-term community-based follow-up study

In 1991–1993, all first-year students in upper secondary school (16–17 years old) in the Swedish university town of Uppsala, with approximately 180,000 inhabitants, were asked to participate in a screening for depression [28]. School dropouts were also invited. Out of a total of 2465 individuals, 93% (n = 2300) participated in the screening, which included two self-evaluations of depression: the Beck Depression Inventory-Child and the Centre for Epidemiological Studies-Depression Scale for Children [29]. Students with high scores and those who reported a suicide attempt were interviewed with the Diagnostic Interview for Children and Adolescents with a revised form according to the DSM-III-R (DICA-R-A) [30]. In all, 355 students in the screening were classified as suffering from depression and were accordingly selected for a diagnostic interview. For each depressed student, a same-sex class-mate and with low scores in the screening was recruited into a comparison group. In total, 609 individuals (n = 307 in the depressed group and n = 302 in the control group) participated in the diagnostic interview and consented to be contacted for a future follow-up study. At the time of the interview, they also completed a range of self-rating measures, including the Somatic Symptom Checklist Instrument (SCI) on somatic symptoms. Some of the participants in the comparison group (n = 65) were retrospectively diagnosed with major depression or dysthymia occurring before the baseline study and consequently were included in the depression group. Some of the participants with positive screenings did not meet the criteria for a depressive disorder upon being interviewed for current and lifetime major depression or dysthymia and were in the present analyses relocated to the control group (n = 55). Approximately 15 years after the baseline study, the participants who had consented to a follow-up study were contacted and invited to a follow-up interview. They were also asked if they wanted to participate in studies that included health registers. Data were subsequently collected from health registers 17–19 years after the baseline study. Among the 609 individuals who had participated in the diagnostic interview and who also had completed the SCI at baseline, approximately 70% participated in the follow-up interview. Of these, 375 individuals gave their written consent to be followed through the health registers (n = 182 in the depression group and n = 193 in the control group). The procedure is outlined in Fig. 1. Further information about the follow-up study is provided elsewhere [27, 31].

Adolescent depression was defined as major depressive disorder (MDD) or dysthymia according to DICA-R-A [30] (see Fig. 1).

Adolescent anxiety was defined as any anxiety disorder according to DICA-R-A [30].

The SCI is a Swedish version of the Psychosomatic Symptom Checklist [32]. The SCI assesses 22 items reflecting various somatic symptoms: tiredness, headache, feeling chilly, insomnia, eye tiredness, abdominal pain, dizziness, nausea, perspiration, appetite problem, breathing problem, polyuria, limb pain, itching, dry mouth, palpitation, constipation, fainting, regurgitation, chewing pain, and swallowing problems. Allergy was part of the checklist but was excluded from the analyses because it was considered a somatic disease rather than a symptom. The symptoms were graded in frequency (0 = never, 1 = monthly, 2 = weekly, 3 = several times a week, and 4 = daily), and intensity (0 = no problem, 1 = minor, 2 = moderate, 3 = troublesome, and 4 = extremely troublesome), for the last month. The questionnaire has been used in previous publications [12, 23, 27]. A somatic symptom was recorded when the frequency and intensity were multiplied to yield a score ≥ 6 (e.g., 2 × 3: weekly × troublesome symptoms). Such a scoring approach excluded minor problems and the possibility that monthly premenstrual symptoms would be recorded as positive. The same cut-off has been used in earlier publications [23, 27].

In the analyses of the control group, somatic symptoms were categorized as 0 vs. ≥ 1 symptoms (a more fine-grained categorization was not possible due to small numbers in the cells). In the analyses of individuals with adolescent depression, four categories of somatic symptoms were created: 0, 1, 2–4, and ≥ 5 symptoms—a categorization that was grounded in our previous analyses of the same data material, where ≥ 5 somatic symptoms were found to characterize a threshold value in the prediction of mental health outcomes in adulthood [27].

A set of potential confounders, which may potentially have affected both somatic symptoms at baseline and mental disorders in adulthood, were used to adjust the analyses. Information on conflicts between parents, conflicts with parents, economic hardship, parental unemployment, and somatic illness collected at baseline through the Children’s Life Inventory [33] was included. In addition, we included information on physical/sexual abuse in childhood collected retrospectively in the follow-up study [31]. Conflicts between parents and conflicts with parents were shown to be significantly related to major depression at baseline [34] as well as to somatic symptoms at baseline [23], and analyses of the follow-up data demonstrated that conflicts with parents and physical/sexual abuse in childhood were associated with mental disorders in adulthood [31]. Socioeconomic status in the family of origin had been shown to be associated with mental disorders in adulthood [35]; therefore, measures of economic hardship and parental unemployment collected at baseline were included. In order to account for the fact that some somatic symptoms might have had a medical explanation (i.e. due to somatic illness), a measure of somatic illness reported in adolescence was included. The variable was created from two items from the Children’s Life Inventory [33]: “I have been severely ill or injured”, and “I have been hospitalized more than one week”, with the possible response categories “During the past year” and “Earlier in life”. The measure of somatic illness was defined by a positive record on at least one of these two items, i.e., self-reported somatic illness or injury some time in life until baseline and/or the adolescent’s report on having ever been hospitalized more than 1 week some time in life until baseline. The data did however not include any information about specific somatic diagnoses.

The Swedish National Health and Welfare Board maintains the official registers concerning health and sickness in Sweden. The national patient register was used in the present study from 1992 until 2009. The national patient register includes data on inpatient care and outpatient hospital-based care. With regard to inpatient care, the register data cover almost all inpatient visits since 1987. With regard to hospital-based outpatient care, outpatient visits have been registered since 2001, but only a part of the data is covered during the follow-up period. Hospital-based mental health care diagnoses were classified according to ICD-10 criteria—specifically, the codes F10–F69 were used to define hospital-based mental health care. For more detailed analyses, the diagnoses were also divided into different general categories: F10–F19, mental and behavioral disorders due to psychoactive substance use; F20–F29, schizophrenia, schizotypal and delusional disorders; F30–F39, mood disorders; F40–F48, neurotic, stress-related and somatoform disorders (including all anxiety disorders); F50–F59, behavioral syndromes associated with physiological disturbances and physical factors; and F60–F69, disorders of adult personality and behavior.

Binary logistic regression analyses were performed to assess the association of somatic symptoms in adolescence with later hospital-based mental health care. Adjustments were made for adolescent depression and anxiety, sex and other potential confounders. Odds ratios with 95% confidence intervals were reported. In the descriptive analyses of somatic symptoms and specific mental health care diagnoses, when several categories of somatic symptoms were compared, linear-by-linear associations were used to calculate linear relationships. To compare the groups of individuals with 0 and ≥ 1 somatic symptoms at baseline, respectively, the Fisher’s exact test was used. Stata version 15 (StataCorp, College Station, TX) was used.

The data and materials had several strengths. The baseline data were population-based, including 2300 adolescents of the same age, with a high participation rate (93%) in the depression screening. Another advantage was the long follow-up period from adolescence to adulthood. The prospective study design and the use of register data enabled us to follow individuals over time and to avoid the problem of recall bias. The data also provided the opportunity to investigate mental disorders and somatic symptoms at both baseline and follow-up (although neurotic, stress-related and somatoform disorders at follow-up were grouped together). A limitation was that only about two-thirds of participants in the original investigation were included in the present register-based follow-up. Yet, the participation rate can be seen as reasonably high in relation to the follow-up period. Furthermore, the attrition rates at follow-up were similar between the depressed and control groups. We assessed bivariate associations between somatic symptoms and later hospital-based mental health care in the two groups separately. To investigate the prediction of somatic symptoms whilst also including a set of potential confounders, we also performed analyses of the pooled sample. This design has limitations since the groups of depressed adolescents and their non-depressed matched peers were different in several respects, as shown in Table 1. Since only a fraction of non-depressed adolescents were included in the data, the pooled sample is not representative of the original population of 16–17-year-olds in the city of Uppsala. Yet, when assessing the relationship between adolescent somatic symptoms and later hospital-based mental health care, it is of high relevance to control not only for adolescent depression but also for anxiety and other confounders and in this study, this required a pooled sample.

In the present study, we chose to focus on severe mental illness and not on total consumption of mental health care. We did not use information about psychological and pharmacological treatment of mental disorders in general practitioner care, despite the fact that most patients with mental health conditions in Sweden are treated by a general practitioner [45]. Such information could have been of value. A limitation with the strategy of focusing on hospital-based mental health care is also that the actual number of participants who receive such specialized care is relatively small. Another limitation is that a major proportion of adults suffering from mental disorders does not seek or receive adequate treatment. Help-seeking behavior is lower among men than among women, and untreated mental disorders are not uncommon [46]. Hence, it is likely that there are individuals captured in our data who suffer from severe mental disorders without having received hospital-based treatment. This might result in an underestimation of the actual need of adult hospital-based care. Furthermore, the data on hospital-based outpatient care did not include all registered cases, which implies an underestimation of the total use of hospital-based care and a higher weight of in-patient care compared to out-patient care. Still it seems unlikely that the general findings in relation to our research questions would be affected.

Finally, we lack data on specific somatic diagnoses in adolescence. Hence, we were not able to disentangle whether the association between somatic symptoms in adolescence and hospital-based mental health care in adulthood was due to somatic symptoms with or without a medical explanation. While we did include a measure of hospitalization due to somatic illness or injury in adolescence, this variable might have captured only a portion of the adolescents with somatic illness. Another limitation with this measure is that it was based on adolescents’ self-reports.